Enterprise AI, Data & Platform Engineering. 30+ years modernizing financial-services technology across mortgage, banking, risk, compliance, and treasury — now architecting governed AI reasoning systems for Fannie Mae Multifamily.
“The model is the electricity. The harness is the grid.” — Signal Over Noise, weekly Agentic AI briefing
Technology architecture and engineering executive with 30+ years of experience leading financial-services modernization across mortgage, banking, risk, compliance, treasury, regulatory reporting, and enterprise data platforms.
Currently Senior Director of Technology Architecture & Engineering for Fannie Mae Multifamily, leading next-generation AI, data, and cloud-modernization strategy. Recognized for building governed AI reasoning systems that combine generative AI, knowledge graphs, vector search, deterministic validation, and human oversight with reusable platform patterns.
Public builder and thought leader with a 60-repository open-source reference architecture portfolio and Signal Over Noise, a weekly Agentic AI briefing platform.
Knowledge graphs, GraphRAG/RAG, vector search, knowledge engineering, data modernization, unstructured-data strategy.
API strategy, event-driven systems, composable architecture, serverless, microservices, platform modernization.
Converts emerging technology concepts into pilots, reusable frameworks, production platforms, operating models, and delivery acceleration mechanisms.
Audit trails, lineage, telemetry, model validation, AI safety, compliance, security, and enterprise-grade reliability.
Partners with senior leaders to align platform vision, adoption roadmaps, architecture principles, and investment priorities across complex portfolios.
Builds AI literacy, mentors architects and engineers, publishes AI insights, and open-sources practical reference implementations.
Defined and advanced target-state AI, data, and reasoning architecture for Fannie Mae Multifamily, moving the organization from isolated AI experiments toward a governed, platform-centric ecosystem.
Pioneered a System of Reasoning architecture integrating GenAI, knowledge graphs, deterministic rules, lineage, telemetry, and human oversight to support explainable, decision-aware financial-services workflows.
Led hybrid graph + vector architecture patterns for document intelligence, underwriting, surveillance, semantic search, natural-language access, and API-based integration across structured and unstructured data.
Delivered reusable GenAI and knowledge-graph capabilities that transform unstructured documents into structured, queryable knowledge assets and significantly reduce manual analysis effort.
Championed GenAI-powered delivery acceleration and agentic automation, including vulnerability-remediation workflows that compress manual cycles from days to minutes.
Modernized major banking and mortgage technology portfolios across Fannie Mae, Capital One, and KPMG — spanning risk, AML, compliance, regulatory reporting, loan sourcing, smart contracts, and AI-enabled document automation.
Lead technology architecture and engineering strategy for Multifamily modernization, with emphasis on enterprise AI, data architecture, cloud-native platforms, knowledge graphs, security, compliance, and cost-aware delivery.
Provided technical leadership and modernization vision for the financial backend of Capital One, with focus on cloud-native architecture, machine learning, agility, scalability, and performance.
A newsletter and content platform, created and published by Srini, examining the technical, architectural, and operational realities of running autonomous AI agents in production — architecture patterns, testing, security, cost, and the shift from DevOps to AgentOps.
Rather than celebrating capability announcements, it treats agentic AI as an engineering discipline: what breaks in production, what governance actually requires, and what a durable harness around a model looks like.
“The model is the electricity. The harness is the grid.”
Architecture patterns for multi-agent systems, loop engineering, and interface design; production concerns spanning testing, security, and cost optimization.
Deep dives on specific implementations, frameworks, and infrastructure evolution from DevOps to AgentOps.
Visual ecosystem maps and decision matrices — e.g. mapping the Google GenAI stack.
Podcast and video companions extending each week's brief beyond the written issue — on Apple Podcasts and YouTube.
A 60-repository public GitHub portfolio spanning Generative AI, Agentic AI, MCP, knowledge graphs, document intelligence, serverless, blockchain, and data engineering.
A suite of Model Context Protocol servers that lets an LLM orchestrate specialized services — flights, hotels, events, weather, and budget analysis — into a single itinerary.
github.com/skarlekar/mcp_travelassistant →An LLM-powered knowledge base with an interactive graph viewer. Drop raw sources in, and it writes and maintains structured, cross-linked wiki pages — summaries, concepts, entities, and synthesis.
github.com/skarlekar/knowledge-compiler →Intelligent document processing that uses LLMs and agentic AI to extract structured knowledge from unstructured documents — auto-generating ontologies and entity/relationship graphs.
github.com/skarlekar/deepinsight →Constructs knowledge graphs from unstructured documents and exposes them to natural-language querying — the pattern behind graph-based document intelligence.
github.com/skarlekar/doc2graph →A chatbot trained on a corpus of user-provided documents — an early reference implementation of retrieval-grounded subject-matter expertise.
github.com/skarlekar/digital-sme →A prompt storehouse — a library, generation tools, and comparison utilities for building and managing AI prompts under open collaboration.
github.com/skarlekar/prompt-works →An AI image-generation playground powered by Gemini — "conjure something beautiful" as an interactive creative-AI demo.
github.com/skarlekar/mybanana →A Quora-like decentralized knowledge-sharing app built on Hyperledger Fabric, with smart contracts incentivizing quality content and OAuth-based identity.
github.com/skarlekar/smart-quora →George Mason University
University of Oxford
Madurai Kamaraj University
Madras University